A system with multiple transmitters and multiple receivers (Multiple Input Multiple Output: MIMO) has been applied to communication systems so as to achieve diversity and/or high rates of data transmission. In such a MIMO system, multiple source signals are multiplexed into one mixed signal and transmitted, and the signal retrieved from a medium consists of a mixture of the signals of interest. A MIMO system is used not only for communication systems but also for recording signals and information to high density recording media.
On a receiving side or a reading side of signal from a recording medium, it is necessary to separate the respective source signals from the mixed signal and retrieve the source signals in the same order as they were transmitted to the medium or recorded on the medium. Conventionally, in such a system, training data is used to estimate channel state information (CSI) or inverse channel state information (ICSI). On the receiver side, an approximate CSI or inverse channel state information (ICSI) is then used to separate the mixed signals or initialize the coefficients of a blind adaptive algorithm, which adaptively adjusts its coefficients, ensuring that the output always consists of de-mixed signals. However, as described later, use of the training signal causes reduction of the throughput of the communication system or reduction of the storage capacity of a high density recording medium.
An approach which is directly related to this patent is a complete blind identification of MIMO CSI or MIMO ICSI [1]-[5]. In these methods, only the received signal is used, and no priori knowledge of the transmitted signals is assumed. Based on this assumption, there are several approaches that have been used to estimate the CSI or ICSI of a MIMO channel. Basically, these estimation methods can be classified as based on: the original signal properties such as orthogonality, constant modulus, and cyclo-stationarity; assumption that the sources are independent; and sub-space based techniques [1]-[12]. In all these methods, separation of the mixed signals is possible. However, since a complete blind identification of CSI has multiple solutions, each of which are permutations of another, it is not possible to assign a separated source signal to a specific source on the transmitter side [8]-[12]. When a solution is given as a sequence, permutation is a sequence in which the order of elements of such a solution are permutated. When the statistics of the source signals are used as a cost function, it is also possible that the blind algorithm will result in a solution where the in-phase or quadrature component of a given source is permutated with an in-phase or quadrature component of another source.
In blind source separation (BSS) of convolutive mixed sources, frequency domain has been utilized to minimize computational complexity that is associated with time domain BSS [13]-[15]. In one example, the signals received by each of the receiving sensors are firstly converted to the frequency domain using FFT (fast Fourier transform). Once this has been done, the convolutive mixture problem is converted to an instantaneous mixture problem of the respective bins of the sources [13], [15]. Furthermore, by utilizing the method explained in [16], it is possible to make permutation that occurs in each frequency bin uniform to all the frequency bins. As an example, FIG. 1 shows a system for BSS where the received signal is processed in the frequency domain.
The system shown in FIG. 1 is provided with: MIMO medium 101 which takes signals x1, x2 as input and signals y1, y2 as output; FFT operation unit 102 performing N-point FFT on signal y1; FFT operation unit 103 performing N-point FFT on signal y2; blind separation operation unit 104 applying the blind signal separation algorithm to the outputs of these FFT operation units 102, 103; and IFFT operation units 105, 106 performing N-point IFFT (inverse fast Fourier transform) on the output of blind separation operation unit 104 to output signals z1, z2, respectively. In this system:
(a) x1 and x2 are the original source signals, with a Fourier transform {X1,1 . . . X1,NF} and {X2,1 . . . X2,NF}, respectively;
(b) y1 and y2 are the received signals after passing through MIMO transmission medium 101. {Y1,1 . . . Y1,N}, which are supplied from FFT operation unit 102, are the frequency bins of the Fourier transform of the received signal y1. Similarly, {Y2,1 . . . Y2,N}, which are supplied from FFT operation unit 103, corresponds to N pieces of the frequency bins of the Fourier transform of received signal y2; and
(c) After pairing the frequency bins of the received signals into pairs s1 to sN, a blind separation algorithm is used by blind separation operation unit 104 to separate the bins corresponding to the original signals x1 and x2, respectively. Then, signals z1, z2 are obtained by applying inverse Fourier transform by IFFT operation units 105, 106. In case there is no permutation of bins as specified in reference [16], the frequency bins Z1,1 to Z1,N will correspond to only one of x1 and x2, and signal z1 will thus corresponds to only one of x1 and x2. On the other hand, if there is a permutation, the first bin Z1,1 may correspond to X1,1, while the next bin Z1,2 may correspond to X2,2. Thus, though it is possible to separate the sources without their frequency components being permutated, the problem of permutation of the sources still remains.
The references cited in this description will be listed below:
[1] Chong-Yung Chi; Chii-Horng Chen, “Cumulant-based inverse filter criteria for MIMO blind deconvolution: properties, algorithms, and application to DS/CDMA systems in multipath,” IEEE Transactions on Signal Processing, Volume: 49, Issue: 7, Pages: 1282-1299, July 2001
[2] Chor Tin Ma; Zhi Ding; Sze Fong Yau, “A two-stage algorithm for MIMO blind deconvolution of nonstationary colored signals,” IEEE Transactions on Signal Processing, Volume: 48, Issue: 4, Pages: 1187-1192, April 2000
[3] Ka Lok Yeung; Sze Fong Yau, “A super-exponential algorithm for blind deconvolution of MIMO system,” Circuits and Systems, 1997. ISCAS '97, Volume: 4, Pages: 2517-2520, 9-12 Jun. 1997
[4] Comon, P.; Moreau, E., “Blind MIMO equalization and joint-diagonalization criteria,” Acoustics, Speech, and Signal Processing, 2001. Proceedings. (ICASSP '01), Volume: 5, Pages: 2749-2752, 7-11 May 2001
[5] Kung, S. Y.; Xinying Zhang, “An associative memory approach to blind signal recovery for SIMO/MIMO systems,” Neural Networks for Signal Processing XI, 2001. Proceedings of the 2001 IEEE Signal Processing Society Workshop, Pages: 343-362, 10-12 Sep. 2001
[6] T. L. Marzetta, “BLAST Training: Estimating Channel Characteristics for High Capacity Space-Time Wireless,” Proc. 37th Annual Allerton Conference on Communication, Control, and Computing, pp. 958-966, September 1999
[7] Li Yuanjie and L. Yang, “Semi-blind mimo channel identification based on error adjustment,” 2003. Proceedings of the 2003 International Conference on Neural Networks and Signal Processing, Vol. 2, pp. 1429-1432, Dec. 14-17, 2003
[8] Y. Li, K. J. R. Liu, “Adaptive Blind Source Separation for Multiple-Input/Multiple-Output Systems,” IEEE Trans., Inform., Theory, vol. 44, No. 7, pp. 2864-2876, November 1998
[9] B. G. Agee, “Blind separation and capture of communication signals using a multitarget constant modulus beamformer,” MILCOM '89. Conference Record. vol. 2, pp. 340-346, October 1989
[10] P. Sansrimahachai, D. B. Ward, A. G. Constantinides, “Multiple-input multiple-output least-squares constant modulus algorithms,” GLOBECOM '03. IEEE, vol. 4, pp. 2084-2088, December 2003
[11] J. J. Shynk, R. P. Gooch, “Performance analysis of the multistage CMA adaptive beamformer,” MILCOM '94. vol. 2, pp. 316-320, 2-5 Oct. 1994
[12] M. Hajian, J. Fernandes, P. Cunha, L. P. Ligthart, “Adaptive equalization for mobile communication systems based on constant modulus algorithm,” Vehicular Technology Conference, 2001. IEEE VTS 53rd, vol. 3, pp. 1609-1613, May 2001
[13] Belouchrani, A.; Amin, M G, “Blind source separation based on time-frequency signal representations,” Signal Processing, IEEE Transactions on, Volume: 46, Issue: 11, Pages: 2888-2897, November 1998
[14] M. Z. Ikram and D. R. Morgan, “A beamforming approach to permutation alignment for multichannel frequency-domain blind speech separation,” Proc. ICASSP, pp. 881-884, May 2002
[15] T. W. Lee, A. J. Bell, and R. Orgimeister, “Blind source separation of real world signals,” Proc. ICNN, pp. 2129-2135, June 1997
[16] H. Sawada, R. Mukai, S. Araki and S. Makino, “A Robust and Precise Method for Solving the Permutation Problem of Frequency-Domain Blind Source Separation,” IEEE Trans., Speech Audio Processing, vol. 12, no. 5, Sep. 2004
[17] D. Godard, “Self-Recovering Equalization and Carrier Tracking in Two-Dimensional Data Communication Systems,” IEEE Transactions on Communications, vol. 28, issue: 11, pp. 1-867-1875, November 1980
[18] K. Oh and Y. Chin, “Modified constant modulus algorithm: Blind equalization and carrier phase recovery algorithm,” Proc. IEEE ICC, vol. 1, pp. 498-502, June 1995
[19] Van der Veen, A.-J., “Statistical performance analysis of the algebraic constant modulus algorithm,” Signal Processing, IEEE Transactions on, Volume: 50, Issue: 12, Pages: 3083-3097, December 2002
[20] Regalia, P. A., “A finite-interval constant modulus algorithm,” Acoustics, Speech, and Signal Processing, 2002. Proceedings. (ICASSP '02). IEEE International Conference on, Volume: 3, Pages: III-2285-III-2288, 13-17 May 2002